Ever-increasing amounts of data are driving the military to look to automated ways to detect patterns and threats.

The Air Force, for example, says current methods of making computer models that support automated reasoning and learning require painstaking manual labor. But reality isn't static, meaning that new relevant information may come in that the model can't accommodate. Plus, the model subject--the "domain of interest" the model seeks to examine--itself will change over time. But, even small changes to a model can be disruptive. What's needed, according to an Air Force broad agency announcement updated Oct. 13, are new algorithms that allow computer models support uncertainty and inconsistency.

The Defense Advance Research Project Agency, meanwhile recently made a broad, agency announcement for the automatic detection of threats. The military today relies on roomfuls of operators who fuse multi-sensor data by literally talking (or instant messaging) about what they're seeing.

Were an automatic method for correlating data with threats to exist, operators would be more efficient and ensure the sensors themselves are better utilized, the DARPA announcement states.

For more:
- see the Air Force broad agency announcement or the DARPA announcement

Related Articles:
DARPA in pursuit of insider threats
Robot farmers and psychiatrists coming in a not-so distant future
IARPA looks to the future